Estimated reading time: 7 minutes

The U.S. Census Bureau serves a vital role in providing statistical information about the nation’s people and economy. Amid the COVID-19 pandemic, we’ve had to adapt our surveys and other data collections to keep both the nation informed, and our staff and the people who respond safe. 

For example, we described how we adapted the 2020 Census in the Adapting Field Operations to Meet Unprecedented Challenges blog.

In this blog, we talk about how the pandemic has disrupted data collection for one of the nation’s most comprehensive sources of information about the U.S. population — the Census Bureau’s American Community Survey (ACS). 

Instead of counting the entire population once a decade like the census, the ACS surveys a sample of the population (about 3.5 million addresses) every year to create statistics for the nation, states and communities down to the “tract” or neighborhood level.

In fact, the ACS launched in 2005 to replace the census “long form.” The census long form asked about one in every six households detailed questions. The remaining households received a “short form” with a pared down set of questions. The long form produced the detailed social, economic and housing characteristics that complemented the national headcount each decade.

However, the Census Bureau and Congress recognized that it wasn’t enough to update statistics once a decade. The nation needed timelier information on communities.

As an annual survey, the ACS is the nation’s most current, reliable and accessible data source for many local statistics. These statistics are comparable across the nation, and many do not exist anywhere else.

The ACS expands beyond the questions on the 2020 Census questionnaire and provides information on critical planning topics such as:

• Educational attainment.
• Household and personal income.
• Veterans status.
• Employment.
• Commuting time.
• Housing costs.
• Disability status.

Combined with the census, ACS data help determine how hundreds of billions of dollars in federal and state funds are distributed each year.

The pandemic significantly disrupted how we collect this critical information. First, we’ll explain how we usually collect the data, and then how we adapted amid the pandemic.

Each month, we ask about 295,000 housing unit addresses to participate in the survey. We call each month’s set of addresses a “panel.”

We collect data for each panel through two phases:

• Self-Response. We first mail each address an invitation to respond online. Later, we send nonresponding addresses a paper questionnaire, giving them the option of responding by mail. Depending on if and when they respond, an address may receive up to five mailings encouraging them to participate.

• Nonresponse Followup. If we don’t receive a response online or by mail, we may send a field representative to visit the address. (If we have a phone number, the field representative tries calling them first.) The field representative uses a laptop to guide them through asking the survey questions and recording the responses.

The pandemic posed challenges for collecting ACS data during both phases. 

While addresses could still self-respond online or by mail amid stay-at-home orders, such restrictions limited our ability to mail questionnaires and invitations to participate.

We rely on our National Processing Center (NPC) in Indiana to print, assemble and mail out survey invitations; capture data from completed forms; and provide telephone support for people who respond.

Local stay-at-home orders in Indiana and the need for social distancing created substantial disruptions to most of this work because it could not be done virtually.

We suspended key activities from mid-March through June 2020, including:

• Mailing invitations, questionnaires and reminders. However, addresses that had already received invitations could continue responding online with no disruption.
• Processing paper questionnaires that households returned by mail. We stored these questionnaires in our secure facility until our staff could return to scan them and process the information.

We ultimately canceled the mailouts for the April, May and June 2020 panels. Instead, our field representatives called a sample of these addresses (when we had a phone number available) to collect the information. These changes to our data collection increase the potential for bias in our estimates. In a future blog, we will detail how these biases can occur and steps we are taking to mitigate the impact of these changes on our 2020 data products.

In early June 2020, employees gradually began returning to the NPC facility. We were still limited by social distancing requirements and staff shortages, but we were able to resume mailing survey materials for the July 2020 panel using a revised strategy.

Instead of our usual five mailings, we sent two that included:

• Information on how to respond online. This first mailing typically included a variety of materials encouraging people living at the address to respond. We first mailed the leftover packages that had been assembled before the stay-at-home orders began. As we needed to print and assemble more, we streamlined what was included.
• Paper questionnaires for some addresses and a reminder to respond online for the others. For this second mailing, we had limited capacity to address the thick questionnaire packages, so we prioritized sending those to areas with low-response or low internet connectivity. However, given that we publish data nationwide, we ensured that paper questionnaires were sent to some addresses in all areas.

In October 2020, as more staff returned to in-person work, we were able to resume:

• Mailing questionnaire packages to all addresses in the current panels that had not yet responded. 
• Sending a third mailing to remind households to respond.
  

We continued doing this until we were able to return to our standard five-mailout strategy in April 2021.

During the pandemic, we’ve been able to continue following up with households that did not respond. However, between mid-March and June 2020, our field representatives did not visit households in person. They only called households (when we had a phone number available).

We resumed visits for some areas of the country in July 2020 and in all areas without stay-at-home orders in September 2020.

Our regional office staff worked closely with each tribal government with land-access restrictions to explain the importance of our surveys, ask permission to conduct the ACS on their tribal lands, and learn the specific protocol our field staff should follow on their lands.

The pandemic also significantly impacted how we collect information from people living in group quarters, such as nursing homes, college dorms and prisons. (This population was a challenge for the 2020 Census, too, as discussed in our 2020 Census Group Quarters blog.)

We typically rely on personal visits to collect data from group quarters, and pandemic-related restrictions temporarily halted those.

In a future blog, we’ll talk more about the impacts on data collection for group quarters, and how we:

• Suspended collecting information from group quarters from April through June 2020.
  
• Gradually resumed collecting data from group quarters primarily by phone in July 2020.
• Have explored using administrative records as a source of information about this population.

The need to protect our staff and the people who respond temporarily shut down many of our operations:

• Mailing information to households encouraging them to respond.
• Following up in person with nonresponding households.
• Collecting data from group quarters.

We were able to adapt to the challenges of the pandemic while still managing to collect about two-thirds of the responses we usually collect in a survey year.

The natural question is: What does this mean for the detailed demographic, social, economic and housing statistics that the ACS will produce this fall?

We expect the 2020 ACS statistics may look different from prior years — both because of what the survey measured about our nation during such a historic year and because of the pandemic’s effects on the survey itself.

In future blogs, we will discuss:

• How we’re adapting our estimation techniques to respond to the data collection challenges, which should improve the resulting statistics.
• What it means for the data products we’ll release in the fall.
• Our assessment of the quality of the 2020 ACS estimates.